Conventionally, among images (still images or moving images) uploaded to SNS (social networking service) by users, there are many images containing product images, each of which is an image of a product. Many images on SNS are associated with, for example, text containing keywords or tags, and thus a search using keywords and the like can be performed (see “Twitter”, accessed on Jul. 28, 2015 at: twitter.com/, for example).
Thus, it is conceivable to accumulate information indicating a relationship between products and consumer attributes, for example, indicating which sex, which age group, which race, or the like more consumers who purchased a certain product belong to, using images on SNS, and to use the accumulated information for, for example, analysis of consumer behavior, market research, and the like.
One method for realizing that processing is a method for analyzing text attached to images. However, the proportion of images associated with text with which a product can be identified, with respect to images containing product images, is not large, and, furthermore, problems such as the language barrier or spam make it difficult to analyze text and accumulate this type of information.
Another method for realizing that processing is a method for analyzing images themselves. Image analysis has raised expectations as a method for accumulating this type of information from many of the images containing product images.